Rawr.Mage - Page 13

globus · 01/30/09, 7:48 AM

Long time reader , first post for me. Sorry for bad english but I'm french.

Rawr is really a great tool that I use daily.
With the 2.1.9 version I think I experiment a bug like Imlach. (I have the same issue in 2.1.8)

here is a snapshot of the sequence reconstruction

I use the near-standard frostfire buid with a 320 seconds fight with standard settings
When I enable sequence reconstruction I see that I consume a mana gem at the beginning of the fight and during the fight I spend a sequence (like 10s) a 0 Mana but with a steady dps. It seems to me that the internal behaviour credits the mana for the first gem (even if I'm at max mana) and then internaly rawr seems to think that I have mana during all the fight.

I understand it may be optimal to stay oom for a few seconds in a fight but for my point of view the dps curve should be near my wand dps (plus some dots) when I'm oom.

even if it's not a bug I would highly appreciate comments/explanations about that, thanks in advance.

Enthorn · 01/30/09, 10:08 AM

Globus, can you check under settings and look at what you have under "Drinking Time"? Is it set to 0?

globus · 01/30/09, 10:51 AM

hello Enthorn,

my drinking time is set to 300.

here is a link to my xml file, so I you want you can investigate more :

RapidShare: Easy Filehosting

only 10 downloads are authorized on this file, hope you'll catch it.

freaknastee · 01/30/09, 12:43 PM

I have a question about RAWR. When I import my character from Armory and set buffs/options with my current spec (56/3/12) I get a result of 4720. Then I goto "Optimize" and it comes up with a score of 4096. How come when I optimize it's giving me a lower score?

Also when you are computing the best spec to use with current available gear (green diamonds), do I have to go through and set glyphs EACH time I want to see how another spec would do, or can I check all glyphs and have RAWR figure it out?

Essentially, I want to input all my available gear + gems with the same fight parameters/options/buffs and compare best spec and gear combination.

The reason I ask is because I'm the only mage in the guild and with arcane my personal DPS is excellent (5200 on Patchwerk, 2'57", 75k rdps). However, I'm concerned about the raid dps gain with my scorch debuff, the guild is melee heavy so I'm trying to think if it will be worth it. (3 locks, 1 boomkin, 1 ele sham, 1 mage)

Raencloud · 01/30/09, 1:40 PM

The DPS is so close between TTW Fire, FFB, and Arcane that in almost any raiding scenario it's going to be better to spec one of the fire specs if you are the only mage. With 5 other casters, it would be a tremendous dps loss to not do so.

Kavan · 01/30/09, 4:18 PM

Originally Posted by Reignman

RapidShare: Easy Filehosting

was also playing around with a guildie who is frostfire... should have the same settings as the above, but fwiw...

RapidShare: Easy Filehosting

I found the problem that was causing this, but unfortunately the fix wasn't ready in time for 2.1.9 release.

Kavan · 01/30/09, 5:11 PM

Originally Posted by globus

I understand it may be optimal to stay oom for a few seconds in a fight but for my point of view the dps curve should be near my wand dps (plus some dots) when I'm oom.

even if it's not a bug I would highly appreciate comments/explanations about that, thanks in advance.

The reason why you get this behavior is because you're not using the advanced solver. Even that one is not perfect and will sometimes have similar problems (a very short segment of oom and ending fight with extra mana) due to its resolution of 30 seconds. One thing I will probably expand in the future will be to give an option to increase the resolution of solver to 10 seconds are maybe even more. For now I just included a warning message if the sequence reconstruction is not successful and you're not using the advanced solver.

That said I'm afraid this won't be of help to you in this case because of an actual bug that I just found out. The advanced solver isn't able to find a solution here and increasing computation limit or switching to depth first won't help (you'll get an out of memory exception). I've identified the cause for the problem, but the solution will probably take some time to implement.

Kavan · 01/30/09, 5:16 PM

Originally Posted by freaknastee

I have a question about RAWR. When I import my character from Armory and set buffs/options with my current spec (56/3/12) I get a result of 4720. Then I goto "Optimize" and it comes up with a score of 4096. How come when I optimize it's giving me a lower score?

Also when you are computing the best spec to use with current available gear (green diamonds), do I have to go through and set glyphs EACH time I want to see how another spec would do, or can I check all glyphs and have RAWR figure it out?

Essentially, I want to input all my available gear + gems with the same fight parameters/options/buffs and compare best spec and gear combination.

For the optimize problem make sure you marked gems and enchants as available.

Optimizer does not currently optimize specs. It will optimize gear for a given spec and fight options. So yes if you would like to see optimal gear for each spec you should set the appropriate glyphs in each case and run optimizer for each spec separately.

Raencloud · 01/30/09, 7:30 PM

Well Kavan, I know how you love your bug reports...

Using optimal arcane gear (non JCing),

Settings - (Total Damage: 1081176.00, 6759.38 score, 6757.35 DPS)
optimal arcane gear w/ best standard enchants + ring enchants and engineering glove enchant
All raid buffs except Mana Spring/Tide totems and no Mixology on Flask of Frost Wyrm
Fight Duration 160 seconds
Focus Magic Uptime set to 1 (also buffed with Focus Magic buff separately from another mage)
No Potion of Wild Magic or Potion of Speed, No Innervate or Mana Tide
Heroism set to optimal
Latency .1
Standard 57/3/11 spec, standard arcane glyphs (molten armor)

The Sequence reads as follows:

00:00.000: Mana Gem (25113 mana)
INFO: Mana Overflow!
00:00.000: Icy Veins+Arcane Power+Mana Gem Effect+ABSpam3MBAM (25113 mana)
00:01.160: Icy Veins+Arcane Power+Mana Gem Effect+AB3ABar3C (24383 mana)
00:12.000: Icy Veins+Arcane Power+AB3Abar3C (20119 mana)
WARNING: Will run out of mana!
INFO: Icy Veins is still up!
00:18.000: AB3ABar3C (17758 mana)
01:54.250: Evocation (0 mana)
02:00.000: Heroism+AB3ABar3C (15780 mana)
02:22.000: Mana Gem (8114 mana)
WARNING: Icy Veins cooldown not ready!
02:22.000: Heroism+Icy Veins+Arcane Power+Mana Gem Effect+ABSpam3MBAM (12895 mana)
02:22.010: Mana Potion (12886 mana)
02:22.010: Heroism+Icy Veins+Arcane Power+Mana Gem Effect+ABSpam3MBAM (17186 mana)

Score 71.78

Lot's of big issues I see in there.
1) It says to do the first rotation for all of 1.160 seconds (that's not even 1 spell, so kind of pointless)
2) The Mana Gem Effect only lasted 12 seconds instead of 15.
3) It tells me Icy Veins is still up, but then shows me not under its effect in the next line 2 seconds early.
4) Evocation spot seems odd, but whatever
5) It warns me that IV is not up, but then uses it anyway 2 seconds before it's up (related to #3?).
6) It pops a mana potion .01 seconds after it tells me to use my cooldowns and only 18 seconds before the fight ends

Additionally,
It seems weird that it's telling me I shouldn't use ABar after MBAM.
With these settings, I'm getting relative stat values of:
1.54 SP
1.35 Int
1.23 Haste
0.84 Crit
Increasing the fight length to only devalues everything, but the rankings all stay in the same place.
The Item Budget also reflects the stat values, yet it still tells me to gem dmg/haste in my pants and shoulders instead of dmg/int.

If you want the XML file, let me know.

*edit, for that matter, I am getting similar results even when in my current armory gear, with int beating out haste 1.21 to 1.15 yet haste being recommended. The spell cast sequence is similar with numbers altered slightly.

Kavan · 01/30/09, 7:48 PM

Originally Posted by Raencloud

If you want the XML file, let me know.

I always want xml file. Without it I can't do anything.

First things first. Did you enable cooldown segmentation and integral mana consumables? If not that probably explains all the problems you're having.

Next to answer some of your questions. You probably have some misconceptions about how the solver works. This is not a discrete simulation, so the concept of single spell casts does not work nicely with it. The sequence reconstruction, as the name implies, tries to reconstruct the sequence based on the solver result. Basically it takes the spell cycles you see in the result and rearranges them in a way that is as close to executable in practice as possible. When it can't achieve that it will give warnings for placements that wouldn't be executable in practice (i.e. cooldowns not ready, oom,...). At the core of the solver is the cycle concept. The cycle characteristics such as dps and mps are computed as accurately as possible (in most cases with a discrete model, in some more complex cases using approximation methods), the solver then uses these cycles under assumption that they are infinitely divisible. This is an assumption that allows the solver to give you the results and as a result some may not be directly translatable to real game play. One could try optimization of the discrete cast model, but that is several orders of magnitude more computation intensive and completely not suitable for gear comparison purposes.

Raencloud · 01/30/09, 8:37 PM

I had integral mana consumables enabled, but not cooldown segmentation.
I also changed the SMP Search Method to DepthFirst, which seemed to help a lot.

What exactly does the score rating indicate? I assume its an indication of how optimal it is. When I was having weird results I was getting a score of 71.78. When I switched to DepthFirst, it went up to 94.5 or so, and when I enabled cooldown segmentation is dropped it to 87.65.

Kavan · 01/30/09, 8:56 PM

Anything below 100 is not executable. 90-100 can usually be tolerated, but lower than that it's more or less garbage. The score is roughly 100 * (1 - sum over each cooldown (duration of unexecutable segments / fight length)) where cooldown includes actual cooldowns and mana. If you can't get a good score with all SMP options enabled it's a good indication that there is a bug involved. Some situations are just incredibly complex to handle (like water elemental with cold snap, which still doesn't work) and it can take me a long time to implement the proper constraints, but first I have to know about the problem before I can fix it.

Raencloud · 01/30/09, 9:08 PM

Rawr XML File is the most recent one use the changes you suggested.

globus · 01/31/09, 4:19 AM

That said I'm afraid this won't be of help to you in this case because of an actual bug that I just found out. The advanced solver isn't able to find a solution here and increasing computation limit or switching to depth first won't help (you'll get an out of memory exception). I've identified the cause for the problem, but the solution will probably take some time to implement.

thanks a lot Kavan for your feedback. It doesn't matter a lot for me if my personal case is not solved immediatly, I post this stange behaviour more in the idea of enhance/correct the greatest soft the mage community have.

May I suggest some enhancement from my point of view :

display directly in the stat panel the hit cap (maybe in the hit score something like 284/288 or in the tooltip).
add a button to automatically add all the stuff of a toon in the available stuff list (great when you import a toon from the armory).
more long term : possibility to store/create several profiles of fight settings and maybe provide default settings for raid boss. great for lazy people to directly choose their boss and have automatically the lenght of fight, the time running/casting...

Astrylian · 01/31/09, 4:42 AM

Originally Posted by globus

display directly in the stat panel the hit cap (maybe in the hit score something like 284/288 or in the tooltip).
add a button to automatically add all the stuff of a toon in the available stuff list (great when you import a toon from the armory).

Those two already exist.